An easy way to visualize the agent ?

[MoMeAKR]

Hello,

I'm looking for a way to actually display what the agent does once trained. I am a bit lost in the MuzeroEvaluator class for instance. Could you provide guidance so I can try and implement some env.render there ? I'll submit a pull request if successful afterwards.
Thanks a lot !

[puyuan1996]

Hello,

Thank you very much for your interest in our work. We are currently organizing the relevant code for rendering and saving replay, and we plan to submit a Pull Request within the next few days. We appreciate your patience.

Best wishes.

[puyuan1996]

Hello,

The example code for loading a trained model and saving the replay is now available in this Pull Request. To store the replay, you can set main_config.env.save_video = True in this file, and specify the path to your trained model's checkpoint file in model_path.

Best wishes.

[MoMe36]

Hello ! Thanks a lot :D I'll be testing ASAP ! Thank you for your work

[puyuan1996]

Hello,

The Pull Request related to real-time video rendering and saving replay has been merged into the main branch. Please refer to the instructions in this file for visualizing the agent.

Best regards.